We Read, We Tweet

“We Read, We Tweet” geographically visualizes Tweets about New York Times articles. Each line connects the location of a tweet to the contextual location of the article it references. The lines are generated based on the sequence in which the tweets occurred.

Personal Blog Writeup

Through visualizing the relationships between a New York Times article’s contextual location, and location of the Twitter users that tweet about the article, a global interest emerges in many stories that pertain to a specific part of the world. I’m fascinated by maps, specifically how boundaries form over time and are consistently being remapped based on regional issues.  Since Twitter has become a prolific tool for disseminating information, yet is so successful because of it’s ephemeral and mobile nature, I wanted to explore how where individual interests lie across the world about articles in the New York Times.

This project involved extensive backend and frontend programming.  The actual data consists of geocoded Tweets and New York Times articles that are stored in a database. Every 10 minutes, a PHP script is run on my server that queries the Backtweets API for any tweets that have occurred regarding specific New York Times articles. For each of the returned tweets, the twitter user’s location is retrieved, and if a valid value is found (there is no standardized system for Twitter user’s location information, Br00k1nz is entirely valid) then the value is geocoded, as well as the New York Time’s article’s geo faceted value using the Google Maps API.  All the information is sanitized, and then inserted in a database.

The front end of the project is written in Java (using the processing.core library). Queries to the database are made based on an article that a user wishes to visualize. The latitude, longitude, and text information is then parsed and mapped in the applet. I decided to create the visualizing using openGL, and create parabolas that showed the  relationships between articles and tweets. A timer is set and displayed at the beginning of each scene, which triggers an connection path each time the timer matches a tweet’s time.

When looking at precedents, I was particularly inspired by both Aaron Koblin’s “Flight Patterns” and Jer Thorp’s “Just Landed” visualizations. Koblin’s visualization elegantly maps air traffic patterns. Some of the images in the series show incredibly intricate networks that are formed by air traffic routes, as well as the airports in a region. Jer Thorp’s processing based visualization shows the locations of twitter users and the places that they fly to, cleverly scraped based on the two tweeted words “just landed”. One of the most compelling aspects of this piece is the 3d translation of data, allowing for an exploration into the intricacies of the paths.

When I first started this project, I set out with a few questions informed my process and methodology. These were:

• What does the distribution of Twitter user’s interests about various topics, locations, and sections from the New York Times look like visually?
• Do current issues in the news effect where Twitter users decide to tweet about?
• Do patterns emerge based on country/region, or are the Tweet/Articles relationships random?
• Are there unseen political/economic/social relationships between countries/regions that are hidden in the data?

After analyzing a few stories, there were a few surprising results.  For instance, an article about Houston electing a gay mayor received mostly tweets from people in the United States. However, another article about the Northwest Airlines terrorist attempt in Detroit attracted tweets from all around the world, the very first actually came from Europe.

Overall, I feel the project was a success. Although it may be hard to discern extensive information as to why person in one location may tweet about another, within a large set of data, it’s quite amazing to see how twitter has been globally espoused, and how diverse many of the tweets are in terms of location.